Download section 4: case studies

SECTION 4: CASE STUDIES
4.1: Introduction
This section presents case studies of Treasury’s forecasting performance. The case studies relate to the
most recent challenges confronting forecasters, namely Mining Boom Mark I (2003-04 to 2007-08)
and the global financial crisis (GFC) and Mining Boom Mark II (2008-09 to 2011-12).
Treasury underestimated the extent of the rise in the terms of trade during Mining Boom Mark I,
which led to the underestimation of nominal economic outcomes and, in turn, taxation revenue, over
the period. In contrast, during the GFC and Mining Boom Mark II, taxation revenue growth has
generally been overestimated, with taxation receipts having struggled to recover from their post-crisis
lows, as accumulated losses continued to flow through the tax system.
The Budget forecast errors (forecast minus actuals) over these periods for nominal GDP and taxation
revenue growth are shown in Figure 4.1, below.
Figure 4.1: Budget Forecast Errors for Nominal GDP and Taxation Revenue Growth
8
Percentage points
Percentage points
Mining Boom Mark I
8
GFC and Mining Boom Mark II
6
6
4
4
2
2
0
0
-2
-2
Non-farm nominal GDP
-4
Total tax revenue
-6
-4
-6
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
2011-12
As part of Treasury’s process of continuous internal evaluation, these forecasting errors have resulted
in several improvements to economic and revenue forecasting methodology and process.
4.2: Mining Boom Mark I
Summary
•
In common with many other forecasters, Treasury underestimated the extent of the rise in
Australia’s terms of trade during Mining Boom Mark I, which led to the underestimation of
nominal economic outcomes. This reflected misjudgement of both demand and supply-side
developments in the mining sector.
•
On the demand side, economic growth in China consistently exceeded forecasts over this period,
on average by around 2½ percentage points per annum. As a result, the demand for steel and, in
turn, iron ore and metallurgical coal, was underestimated.
•
Treasury also overestimated the speed at which global mining production would respond to the rise
in mining output prices. Treasury’s assessment was largely based on the mining companies’ global
43
Review of Treasury Macroeconomic and Revenue Forecasting
projections for production and export volumes, which consistently exceeded actual outcomes. In
Australia this partly reflected infrastructure bottlenecks and the impact of natural disasters.
•
In response to these forecast errors, Treasury has progressed several initiatives designed to
enhance its capability to forecast commodity prices and volumes. As part of the JEFG process, a
balance of payments subcommittee has been established to focus specifically on the outlook for
bulk commodities, comprising Treasury, the RBA, BREE and the ABS. Treasury’s business
liaison program has been refocused to place greater weight on the mining industry. A major project
is also underway which uses detailed projections of commodity supply and demand to model the
outlook for commodity prices through the medium-term.
•
Taxation revenue outcomes were also underestimated, largely reflecting the underestimation of the
strength of the nominal economy during this period. Counterfactual analysis finds that the revenue
forecast errors would have been small if the actual economic outcomes had been known at the time
that the revenue forecasts were being prepared.
•
Capital gains tax (CGT) was a major source of forecast error during Mining Boom Mark I, in large
part, due to the underestimation of the sharp rise in asset prices over the period.
•
In response to these forecast errors, Treasury overhauled its forecasting models to take better
account of the volatile nature of this head of revenue. CGT is now explicitly forecast in a stock
model (rather than being grown in line with other heads of revenue), which better allows for
periods of rapid asset price growth and the accumulation of capital gains and losses.
Background
The first phase of the mining boom spanned the five-year period to 2007-08, with rapid rates of
industrialisation in Asia, particularly in China, increasing worldwide demand for natural resources
and in turn, underpinning an investment boom and a sharp sustained rise in output prices in the
mining sector. As a result, Australia’s terms of trade rose by almost 50 per cent from early 2004 until
the onset of the GFC in 2008. The economy began to approach capacity constraints around 2006 and
2007, with unemployment heading down towards 4 per cent — a rate not seen since the early–1970s.
Monetary policy was tightened to address inflationary pressures, with inflation peaking at 5 per cent
in mid-2008.
Macroeconomic forecasting performance
In common with many other forecasters, a feature of Treasury’s macroeconomic forecasts during
Mining Boom Mark I was the underestimation of the sharp rise in Australia’s terms of trade, and in
turn nominal GDP growth, over the period. The main driver of the rise in the terms of trade, and
hence the main driver of Treasury’s forecast error, was a sustained increase in iron ore and coal
prices. Both demand and supply-side surprises explain the forecast errors. Figure 4.2 highlights the
consistent underestimation of the rise in the terms of trade over successive Budgets.
44
Section 4: Case Studies
Figure 4.2: Budget Forecasts of Australia’s Terms of Trade
200
Index (average pre-2003-04 = 100)
Index (average pre-2003-04 = 100)
200
2012-13
2011-12
2008-09
180
180
2010-11
160
160
2009-10
140
140
2007-08
120
120
2005-06
2006-07
100
100
(f)
80
1993-94
1998-99
2003-04
80
2013-14
2008-09
Note: The figure plots the forecast level of the terms of trade against outcomes, with the forecast level derived from forecast
growth rates of the terms of trade from successive Budgets beginning with the 2005-06 Budget.
On the demand side, forecasters consistently underestimated economic growth in China and other
emerging economies and hence underestimated the demand for steel and in turn iron ore and
metallurgical coal. This pattern was observed over the period of the first mining boom, with forecasts
of the Chinese economy only recently becoming more accurate. Over the period 2003 to 2007, the
average forecast error was 2½ percentage points per annum. Figure 4.3 shows the Consensus
Economics two-year forecasts for Chinese real GDP growth against outcomes.
Figure 4.3: Consensus Economics Two-year Forecasts for Chinese Real GDP growth
16 Per cent
Per cent
14
16
14
2008-09
12
12
Outcom e
2007-08
2011-12
10
10
2006-07
2010-11
8
2004-05
6
1992
1996
2000
2004
2005-06
2009-10
2008
8
6
2012
Source: Consensus Economics and Treasury estimates
On the supply side, despite strong growth in mining investment in response to the sharp rise in
commodity prices, growth in mining production and in turn export volumes was consistently slower
than expected. As early as the 2005-06 Budget, the expectation was that production would quickly
catch-up with forecast demand. The expected supply response was based on what was perceived at the
time to be a conservative assessment of mining companies’ global projections for production and
export volumes, which consistently exceeded actual outcomes. In Australia this has partly reflected
45
Review of Treasury Macroeconomic and Revenue Forecasting
infrastructure bottlenecks and the impact of natural disasters. Figure 4.4 shows that non-rural
commodity export volume forecasts have been consistently overstated, with the error largely due to
overestimation of iron ore and coal export volumes.
Figure 4.4: Evolution of Non-Rural Commodity Export Price and Volume Growth Forecasts
Non-rural commodity export prices
60
Per cent
Per cent
40
60
40
20
20
0
-20
Non-rural commodity export volumes
0
June 2012 outcome
Forecast evolution
Budget forecast
-40
-20
14
14
12
12
10
10
8
8
6
6
4
4
2
2
0
0
-2
-4
-40
Per cent
Per cent
-6
June 2012 outcome
Forecast evolution
Budget forecast
-2
-4
-6
As a result of this experience Treasury has progressed several initiatives to build a deeper
understanding of developments in China and other emerging market economies and the links these
developments have with Australia. In particular:
•
a dedicated unit has been established within Treasury that focusses on the Chinese economy
and Treasury representation in Asia has been enhanced by opening up a new post in India;
•
there has been a greater focus on developing resource industry contacts as part of Treasury’s
business liaison program;
•
a BOP subcommittee meeting has been incorporated into the JEFG process, drawing upon the
expertise of the Bureau of Resources and Energy Economics to forecast the outlook for bulk
commodities;
•
more resources have been devoted to forecasting non-rural commodity production, export
volumes and prices; and
•
a major project is currently underway using industry data to develop a methodology for
projecting bulk non-rural commodity export prices based on a medium-term supply and
demand framework, supplemented by a comparison with projections by BREE, Consensus and
private sector providers.
Revenue forecasting performance
As observed with Treasury’s nominal economy growth forecasts, taxation revenue growth was also
underestimated during Mining Boom Mark I. In particular, nominal economic growth was
underestimated by an average of 1.8 percentage points per annum, compared with an average forecast
error of 2.9 percentage points for taxation revenue growth (Figure 4.5).
46
Section 4: Case Studies
Figure 4.5: Budget Forecast Errors for Growth in Nominal GDP and Taxation Revenue
Percentage points
0
Percentage points
0
-1.8
-2
-2
-2.9
-4
-4
Non-farm nominal GDP errors
Avg GDP error
Tax revenue errors
Avg tax revenue error
-6
-6
2003-04
2004-05
2005-06
2006-07
2007-08
In the case of taxation revenue forecast errors, it is useful to distinguish between those that reflect
errors in the macroeconomic forecasts and those that reflect other sources of error. The latter might
include:
•
error in items that are within scope of the tax base, but not the economic base1;
•
error in estimates of taxation revenue elasticities2;
•
error in estimates of the timing of the receipt of taxation revenue; and
•
miscellaneous factors, such as errors in Budget policy costings, post-Budget policy decisions
and court decisions relating to tax law interpretation.
A counterfactual exercise has been undertaken to decompose taxation revenue forecast errors into
those that reflect macroeconomic forecast errors and those that reflect other sources of error. The
results of this exercise are detailed in Figure 4.6, below. The left-hand panel of Figure 4.6 shows the
actual forecast errors over Mining Boom Mark I. The right-hand panel shows the results of a
counterfactual exercise, which estimates the revenue forecast errors that would have occurred if actual
economic outcomes had been known at the time that the forecasts were being prepared. The presence
of significant revenue forecast error in the counterfactual would suggest scope to improve revenue
forecasting methodology.
The counterfactual exercise assumes in particular that the outcomes for compensation of employees
(the income tax withholding economic base), consumption subject to GST (the GST economic base)
and gross operating surplus (the corporate tax economic base) are known. In the case of capital gains
tax, it is assumed that actual asset price outcomes are known (that is, share prices and house prices).
Not all heads of revenue have been included in this exercise, although the four heads of revenue
considered comprise around 85 per cent of taxation revenue and on average account for around 80 per
cent of the forecast error over the case study period.
1 For example, the company tax base includes deductions for depreciation and current year losses, which are not part of
gross operating surplus (the associated economic base).
2 This is largely an issue for income tax withholding. In contrast, the elasticities of the other heads of revenue to their
taxable bases should be one, by definition, as they have flat tax rates.
47
Review of Treasury Macroeconomic and Revenue Forecasting
Figure 4.6: Contribution to Budget Revenue Forecast Error by Head of Revenue
Actual forecast errors
4
$billion
Counterfactual forecast errors
$billion
$billion
$billion
4
4
2
2
2
0
0
0
0
-2
-2
-2
-2
-4
-4
-4
-4
-6
-6
-6
-8
-8
-8
2
CGT
Income tax withholding
GST
Company tax
Total
-10
-10
2003-04 2004-05 2005-06 2006-07 2007-08
CGT
Income tax withholding
GST
Company tax
Total
-10
4
-6
-8
-10
2003-04 2004-05 2005-06 2006-07 2007-08
The counterfactual exercise finds that most of the taxation revenue forecasting error during Mining
Boom Mark I can be attributed to errors in the nominal economy forecasts. The average actual
revenue forecast error during Mining Boom Mark I was (an underestimate of) $6.2 billion, whereas
the average estimated forecast error under the counterfactual exercise is (an overestimate of)
$0.3 billion. This suggests that the revenue forecast error reflects the underestimation of the strength
of the nominal economy during Mining Boom Mark I, rather than any systematic tendency to
underestimate revenue over this period.
The sections, below, discuss the errors in each of the major heads of revenue over this period.
Income Tax Withholding
Income tax withholding is the largest tax revenue head, comprising over 45 per cent of total taxation
revenue. The relevant economic base for income tax withholding is compensation of employees,
which measures the total wage bill of the economy (and can be split into the total number of
employees multiplied by the average wage). The correlation between the growth in income tax
withholding and the growth in compensation of employees over the past decade (2002-03 to 2011-12)
has been 0.87, indicating a good relationship.3
Figure 4.7 below shows the forecast errors for compensation of employees during Mining Boom
Mark I. The strength in employment growth was consistently underestimated during this period, as
was average wages growth, albeit to a lesser extent. The forecast errors are significantly reduced when
the actual employment and wages outcomes are applied to the income tax withholding model (as
shown in the right-hand panel of Figure 4.6, above), indicating that there was minimal forecasting
error beyond the underestimation of the strength of the labour market over this period.
3 The correlation coefficient has been calculated after adjusting income tax withholding for tax policy changes (largely
personal income tax cuts). Tax policy changes are costed and taken into account separately in the forecasting process.
48
Section 4: Case Studies
Figure 4.7: Budget Forecast Errors for Growth in Compensation of Employees and
Income Tax Withholding
0
Percentage points
-2
Percentage points
0
-2
-2.3
-2.8
-4
-4
-6
Employment errors
Wages errors
ITW error
Comp of employees error
Avg comp of employees error
Avg income tax withholding error
2003-04
2004-05
2005-06
2006-07
-6
2007-08
Company tax
Company tax is the second largest revenue head, comprising over 20 per cent of total revenue. The
main component of the economic base for company tax is gross operating surplus (GOS). The
correlation between the growth in company tax and the growth in GOS over the past decade has been
0.26, indicating that the relationship between the two series is not particularly strong. This is due to
the conceptual differences between corporate taxable income and GOS, and the lag between the
timing of the economic activity and the associated receipt of taxation revenue, which are discussed
further in Section 4.3.
Figure 4.8 shows the Budget forecast errors for GOS and company tax (excluding CGT). The strength
of GOS was generally underestimated during Mining Boom Mark I, which led to the underestimation
of company tax.
Figure 4.8: Budget Forecast Errors for Growth in Gross Operating Surplus and Company Tax
2
Percentage points
Percentage points
0
2
0
-1.4
-2
-2
-1.7
-4
-4
-6
-6
GOS errors
Company tax errors
Avg GOS error
Avg company tax error
-8
-8
2003-04
2004-05
2005-06
2006-07
2007-08
The company tax forecast errors are significantly reduced when the actual GOS outcomes are applied
to the company tax model, (as shown in the right-hand panel of Figure 4.6, above).
49
Review of Treasury Macroeconomic and Revenue Forecasting
Goods and Services Tax
The goods and services tax (GST) comprises around 15 per cent of total taxation revenue. The
primary economic base for GST is consumption subject to GST. The correlation between growth in
GST and consumption subject to GST over the past decade has been 0.75, indicating a good
relationship between the two series.
GST was not a significant source of forecast error during Mining Boom Mark I; however, the head of
revenue was typically underestimated (as shown in the left-hand panel of Figure 4.6, above). When
actual outcomes for consumption subject to GST are applied to the GST model, the forecast errors are
reduced (as shown in the right-hand panel of Figure 4.6 above) and display less bias. This indicates
that there was minimal error in the GST forecasts over this period beyond a slight underestimation of
the strength of consumption subject to GST.
Capital gains tax
Capital gains tax (CGT) is a relatively small, but very volatile, head of revenue. This reflects volatility
in asset prices and variability in both realisation rates (since CGT is only payable when assets are
sold) and also the extent to which capital losses are offset against capital gains. Over the period of the
case study, the annual growth rate of CGT has varied from a rise of 60 per cent in 2004-05 to a fall of
43 per cent in 2009-10, resulting in CGT being a large source of forecast error.
The stock of accrued capital gains accelerated rapidly during Mining Boom Mark I, with asset values
rising at a historically rapid rate. Over this period, CGT was significantly underestimated, particularly
in 2005-06 and 2006-07 (as shown in the left-hand panel of Figure 4.6).
The CGT forecast errors are significantly reduced when the actual outcomes for equity and house
prices are applied to the current CGT model, (as shown in the right-hand panel of Figure 4.6 above)
and the bias in the forecast errors disappears. Treasury does not forecast these asset prices, but instead
adopts the technical assumption that asset prices grow in line with nominal GDP growth. The
counterfactual exercise shows that if the technical assumption had accurately predicted asset prices,
the current CGT model would not have systematically underestimated revenue over this period.
Capital gains tax model improvements
As a result of Treasury’s forecasting experience during Mining Boom Mark I, the CGT models have
been overhauled. The development has been an ongoing process, with the first improvements
implemented in 2006-07, following the sharp rise in CGT revenue in 2005-06.
There are three broad types of CGT taxpayers — individuals, companies and superannuation funds.
The previous methodology did not separately model CGT for individuals and companies. It
effectively assumed that CGT collections would grow in line with the associated taxpayer revenue
stream; for example CGT paid by companies would grow in line with company tax. CGT paid by
superannuation funds was forecast using a simple model based on recent trends and judgement.
The new CGT models are structured to mirror the balance sheet of relevant taxpayers, by tracking the
stock of assets subject to CGT and their price movements. Figure 4.25 (Appendix A) provides an
overview of the model for CGT paid by superannuation funds, by way of an example.
The changes to the CGT modelling framework have improved forecast performance. Figure 4.9 shows
the forecast errors for CGT over the recent period, compared with the errors that would have been
generated using the old CGT forecasting methodology for individuals and companies (which was in
place until 2006-07). The average absolute forecast error for CGT since 2006-07 has been $2.8 billion
50
Section 4: Case Studies
per annum. However, if the old methodology had been in place the average absolute forecast error
would have been much larger at $5.3 billion per annum.
Figure 4.9: Capital Gains Tax Forecast Errors
8
$billion
$billion
Old methodology in place
8
Methodology improvements
6
6
4
4
2
2
-2
-2
-4
-4
CGT forecast errors
-6
CGT forecast errors - old methodology
-8
-6
-8
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
2011-12
4.3: Global Financial Crisis and Mining Boom Mark II
Summary
•
In common with many other forecasters, Treasury did not predict the degeneration of the
US subprime crisis into the GFC in 2008-09 and subsequently overestimated the impact of the
crisis on economic growth in 2009-10.
•
Treasury’s forecasting error for economic growth in 2009-10 largely reflected misjudgements
of the efficacy of Australia’s policy response and the relatively early, and strong, recovery in
economic growth in most of our major trading partners throughout 2009.
•
One lesson from this experience is the importance of having in-house expertise to understand
the operation of financial markets and their linkages with the real economy. To this end the
Treasury has employed a technical specialist with deep financial markets experience and
created a dedicated unit for this purpose.
•
Taxation revenue outcomes were consistently overestimated during this period, despite the
nominal economic forecast errors being close to zero on average. Counterfactual analysis finds
that even if the economic outcomes had been known at the time the forecasts were being
prepared, revenue would still have been overestimated, primarily reflecting company tax errors.
•
In response to the perceived source of these taxation revenue forecasting errors, Treasury is
currently developing a three sector company tax model that takes better account of the different
characteristics of these sectors for taxation purposes, for example, the capital-intensive nature
of the mining sector, and the measurement of finance sector income in the National Accounts.
Background
The world economy experienced a severe financial and economic shock in the second half of 2008.
The Global Financial Crisis (GFC) began in 2007 with the US sub-prime crisis and the crisis
intensified dramatically in September 2008 with the collapse of Lehman Brothers. During this period,
financial conditions deteriorated rapidly, financial and real asset prices collapsed, and business and
51
Review of Treasury Macroeconomic and Revenue Forecasting
consumer confidence fell steeply. The GFC saw the world economy change course sharply from a
five-year period of above-trend growth to the deepest recession since the Great Depression.
The Australian economy performed better than other advanced economies during the GFC. Although
financial conditions were stressed, the financial system held up remarkably well; the economy slowed,
but did not fall into recession; and while unemployment rose, it did so by far less than in many other
advanced economies. The strong performance of the Australian economy largely reflected the strength
of the Australian financial system and public finances; the rapid deployment of fiscal stimulus
measures; the first effects of a significant easing in monetary policy; and a pickup in demand from
China which partly offset pronounced external weakness elsewhere. More broadly, it also reflected
improved policy and institutional arrangements in Australia following a quarter century of reforms
that have made the Australian economy much more resilient to external shocks.4
As the impact of the GFC subsided, Australia’s terms of trade rebounded, with the emergence of a
second phase of the mining boom, and reached a new record high in late 2011 (Mining Boom
Mark II). In contrast to Mining Boom Mark I, the second phase occurred against a backdrop of global
uncertainty and balance sheet consolidation by the corporate and household sectors, reflected in a
significant decline in credit growth. Sectors aligned with mineral resources have grown strongly,
while there has been pressure on some domestic manufacturing and trade-exposed sectors from the
high level of the Australian dollar over this period.
Economic forecasting performance
In common with other forecasters, Treasury did not predict the degeneration of the US subprime crisis
into the GFC in 2008-09 and subsequently overestimated the impact of the crisis on the nominal
economy in 2009-10.
The impact of the GFC on Australia was expected to be severe, with the 2009-10 Budget, prepared in
May 2009, the first in history to forecast a contraction in both real and nominal GDP. Gruen (2010)
notes that at the time the 2009-10 Budget forecasts were being prepared the international context was
dire, with forecasters making significant downgrades to global growth and global trade. For example,
as the global outlook deteriorated, the IMF downgraded its forecasts for 2009 global GDP growth by
4.3 percentage points between October 2008 and April 2009. Over the same period, it downgraded its
forecasts for advanced economy growth by 4.3 percentage points and for global trade volumes by
over 15 percentage points. In line with these forecasters, Treasury significantly downgraded its
forecasts for global growth and correspondingly Australian goods and services export volumes, using
the IMF and OECD forecasts from their inter-country trade models as a guide.
Domestic indicators were also bleak (Gruen, 2010). At the time the 2009-10 Budget forecasts were
being prepared, the latest comprehensive reading on the economy, the December quarter 2008
National Accounts, which was released in early March 2009, showed a contraction of 0.5 per cent. At
the time, this was the weakest quarter — indeed the first negative quarter — since the December
quarter 2000. With subsequent revisions to the National Accounts, that quarter now stands as the
weakest since March 1983. The unemployment rate had also risen sharply from 4.1 per cent in
August 2008 to 5.7 per cent in March 2009.
By the time the Budget forecasts were finalised in April 2009, the average Consensus forecast was
predicting a contraction in Australia’s real economy in 2009 of around 0.6 per cent. As Figure 4.10
shows, the Budget forecast for a contraction of 0.9 per cent was around the median of the range of
forecasts surveyed by Consensus in mid-April.
4 McDonald, T, and Morling, S, 2011, ‘The Australian Economy and the Global Downturn, Part 1: Reasons for resilience’,
Economic Roundup, Issue 2, pp 1-31.
52
Section 4: Case Studies
Figure 4.10: Forecasts for Australian real GDP growth in 2009, as at April 2009
1.6 Per cent
Per cent 1.6
Outcom e
1.2
1.2
0.8
0.8
2009-10 Budget
0.4
0.4
0.0
0.0
-0.4
-0.4
-0.8
Median forecast
-0.8
-1.2
-1.2
-1.6
-1.6
Source: Consensus Economics (Survey date 14 April 2009), Treasury (Budget, 11 May 2009).
Although the real economy did experience a significant slowdown in 2009-10, it performed
significantly better than Treasury expected, largely reflecting misjudgements of the efficacy of
Australia’s policy response and the relatively early, and strong, recovery in economic growth in most
of our major trading partners throughout 2009.5
The largest contributions to the 2009-10 Budget forecast error for real GDP growth were from
household consumption, exports and business investment. Household consumption was forecast to
decline moderately in 2009-10. In the event, it grew by 2.5 per cent. As noted in Gruen (2010),
expansionary monetary and fiscal policy appears to have been large enough and quick enough to
convince consumers and businesses that the domestic slowdown would be relatively mild. In turn this
supported confidence and led consumers and businesses to continue to spend. While the stimulus was
explicitly factored into Treasury’s forecasts, it was also a contributing factor to the 2009-10 forecast
errors.
The 2009-10 Budget forecast exports to contract sharply, by 4 per cent, in 2009-10. In the event, the
outcomes were significantly better than forecast. The outlook for exports was heavily influenced by
expert views on the scale of the global downturn at the time that the forecasts were being prepared.
Within the export forecast error, the main contribution was from non-rural commodity exports, with
smaller contributions from exports of elaborately transformed manufactures and other goods and
services (Figure 4.11). Relative to the 2009-10 Budget forecasts, non-rural commodity exports
performed significantly better than expected, with volumes exceeding forecasts by a significant
margin.
5 Gruen, D, and Stephan, D (2010), ‘Forecasting in the Eye of the Storm’, Address to the NSW Economic Society, 4 June.
53
Review of Treasury Macroeconomic and Revenue Forecasting
Figure 4.11: Contribution to Budget Forecast Errors for Exports in 2009-10, by Component
2
Percentage points
Percentage points
2
0
0
-2
-2
-4
-4
-6
-6
-8
-8
-10
-10
Rural
Non-rural
commodities
ETMs
Other
Total
The better-than-expected outcome for exports largely reflected the better-than-expected performance
of non-Japan Asia, particularly China. Chinese economic activity shifted into more
commodity-intensive sectors, particularly infrastructure spending associated with the Chinese
government’s stimulus packages. There was also substitution away from Chinese domestic production
and towards imports, as lower commodity prices resulted in the closure of some relatively high-cost
Chinese production.
Business investment made the largest contribution to the 2009-10 Budget forecast error. Although,
business investment contracted in 2009-10, the actual contraction was considerably less than forecast.
As Figure 4.12 indicates, all components of business investment fell by less than expected, with
machinery and equipment and non-residential construction finding greater-than-expected support
from the macroeconomic stimulus, and engineering construction benefiting from a rebound in the
outlook for resource commodities reflecting the stronger-than-expected growth performance of
non-Japan Asia, particularly China.
Figure 4.12: Contribution to Budget Forecast Errors for Business Investment in 2009-10,
by Component
2
Percentage points
Percentage points
2
0
0
-2
-2
-4
-4
-6
-6
-8
-8
-10
-10
-12
-12
-14
-14
Non-residential
Engineering
construction
Machinery and
equipment
Other
Total
One lesson from this experience was the importance of having in-house expertise to understand the
operation of financial markets, advances in financial market products and the linkages between
54
Section 4: Case Studies
financial markets and the real economy. To this end, Treasury has employed a technical specialist
with deep financial markets experience and created a dedicated unit for this task.
The Review observes that the large forecasting errors seen during the GFC highlight the fact that
during such extreme events the evolution of the economy is fundamentally less predictable than at
other times. A large number of official agencies overseas also had large forecasting errors during this
period (Figure 4.13). That suggests a case for greater use of risk assessments around the central
forecasts, rather than necessarily a major overhaul of forecasting procedures.
This leads the Review to make the following recommendation:
Recommendation 8:
Scenario analysis is useful as a way of assessing the risks around the economic and revenue
forecasts. Simulation models have an important role to play in this regard and further development
of Treasury’s suite of models may be required to deliver this capability, including in relation to the
international economic outlook.
Figure 4.13: Forecast Errors for Real GDP Growth for Calendar Year 2009,
by Selected Official Agencies Overseas
6
Percentage points
Percentage points
6
Slovenia
Denmark
Japan
Finland
Czech Rep.
UK
Spain
Slovakia
Sweden
Norway
Italy
Canada
-6
Belgium
-6
Austria
-4
US
-4
Switzerland
-2
Portugal
-2
Germany
0
Netherlands
0
Australia
2
Korea
2
Hong Kong
4
Singapore
4
Revenue forecasting performance
The outcomes for the nominal economy and taxation revenue diverged significantly during the GFC,
largely reflecting the impact of accumulated losses on taxable incomes. During the GFC, taxation
revenue fell by 6¼ per cent over 2008-09 and 2009-10, while nominal GDP continued to grow in
year-average terms (Figure 4.14). This resulted in a large decline in the tax-to-GDP ratio.
55
Review of Treasury Macroeconomic and Revenue Forecasting
Figure 4.14: Nominal GDP Growth versus Taxation Revenue Growth
15 Per cent
Per cent 15
Mining Boom Mk I
GFC and Mining Boom Mark II
Nom inal GDP
10
10
Total tax revenue
5
5
0
0
-5
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
-5
2011-12
Treasury’s forecast errors for the nominal economy and taxation revenue also diverged significantly
during the GFC and Mining Boom Mark II. As shown in Figure 4.15, below, the Budget forecasts of
nominal economic growth were underestimated by an average of 0.1 percentage points over this
period, while taxation revenue growth was overestimated by an average of 4.0 percentage points.
Figure 4.15: Budget Forecast Errors for Growth in Nominal GDP and Taxation Revenue
8
Percentage points
Percentage points
6
8
6
4.0
4
4
2
2
0
0
-0.1
-2
Non-farm nominal GDP errors
Tax revenue errors
-4
-4
Avg GDP error
Avg tax revenue error
-6
2008-09
2009-10
2010-11
-2
-6
2011-12
The counterfactual exercise undertaken for Mining Boom Mark I has been repeated for this case
study, with the results detailed in Figure 4.16. The left-hand panel of Figure 4.16 shows the actual
revenue forecast errors and the right-hand panel shows estimates of the forecast errors that would
have occurred if economic outcomes had been known at the time the forecasts were being prepared.
The errors for most heads of revenue are reduced significantly in the counterfactual. However, in
contrast to the Mining Boom Mark I analysis, the overestimation bias in the revenue forecasts does
not disappear in the counterfactual. This is due to company tax overestimation errors, which actually
materially increase in the first two years of the counterfactual. This finding will be examined in more
detail, below.
56
Section 4: Case Studies
Figure 4.16: Contribution to Budget Revenue Forecast Error by Head of Revenue
Actual forecast errors
20
Counterfactual forecast errors
$billion
$billion
$billion
$billion
20
20
15
15
15
10
10
10
10
5
5
5
5
0
0
0
0
-5
-5
-5
CGT
Income tax withholding
GST
Company tax
Total
15
-10
-10
2008-09
2009-10
2010-11
CGT
Income tax withholding
GST
Company tax
Total
-10
2011-12
20
2008-09
2009-10
2010-11
-5
-10
2011-12
The sections, below, discuss the errors in each of the major heads of revenue in turn over this period.
Income Tax Withholding
Income tax withholding was a relatively small source of error during the GFC and Mining Boom
Mark II. The average error over this period for the forecasts of both income tax withholding and
compensation of employees was just -0.1 of a percentage point. When the actual outcomes for
compensation of employees are applied to the income tax withholding model, the forecast errors for
income tax withholding are broadly unchanged (as shown in the right-hand panel of Figure 4.16).
Figure 4.17: Budget Forecast Errors for Growth in Compensation of Employees and
Income Tax Withholding
4 Percentage points
Percentage points
Employment errors
ITW error
Comp of employees error
4
Wages errors
Avg comp of employees error
Avg income tax withholding error
2
2
0
0
-2
-2
-4
-4
2008-09
2009-10
2010-11
2011-12
Company Tax
Company tax made the largest contribution to the overall revenue forecast error in each of the years
since the GFC (left hand panel of Figure 4.16). Company tax revenue is generally a significant source
of forecast error as it is a relatively large, and volatile, head of revenue. This section discusses some
57
Review of Treasury Macroeconomic and Revenue Forecasting
factors that made company tax revenue forecasting difficult over this period. Figure 4.18, below,
shows the forecast errors for GOS and company tax during the GFC and Mining Boom Mark II
period. GOS was underestimated by an average of 1.0 percentage points per annum over this period,
whereas company tax was overestimated by an average of 12.2 percentage points per annum.
Figure 4.18: Budget Forecast Errors for Growth in GOS and Company Tax
20 Percentage points
Percentage points
15
20
15
12.2
10
10
5
5
0
0
-1.0
-5
-5
-10
GOS errors
Company tax errors
Avg GOS error
Avg company tax error
-15
-10
-15
2008-09
2009-10
2010-11
2011-12
The relationship between GOS and company tax is complex. First, there are many conceptual
differences between GOS and corporate taxable income. For example, corporate GOS is measured
before deducting depreciation charges and debt-servicing interest expenses, and excludes holding
gains or losses in trading stock and realised capital gains or losses in the assets and liabilities of the
corporate sector. These conceptual differences are taken into account when translating forecasts of
GOS (blue bars in the left-hand panel of Figure 4.19) into forecasts of company tax on an
income-year basis (red bars in the left-hand panel of Figure 4.19).
Second, there are significant timing differences between these series that reflect the lag between the
earning of economic profits and the receipt of the associated tax revenue. This means that company
tax forecasts on an income year basis (red bars in the left-hand panel of Figure 4.19, below) need to be
translated into company tax forecasts on a cash basis (red bars in the right-hand panel of Figure 4.19),
which introduces a further potential source of forecast error.
Figure 4.24 in the appendix to this section provides more detail about the company tax model.
58
Section 4: Case Studies
Figure 4.19: Company Tax Growth: Forecasts and Outcomes — Income-year and Cash
25
Per cent
Company tax forecast (cash)
Company tax outcome (cash)
25
20
20
20
15
15
15
15
10
10
10
10
5
5
5
5
0
0
0
0
-5
-5
-5
-5
30
Per cent
Per cent
GOS forecast
Company tax forecast (income-year)
GOS outcome
Company tax outcome (income-year)*
25
20
-10
-15
2008-09
2009-10
2010-11
30
Per cent
30
-10
-10
-15
-15
2011-12
30
25
-10
-15
2008-09
2009-10
2010-11
2011-12
Note: Income-year company tax outcomes for 2010-11 or 2011-12 are not yet available.
In both 2008-09 and 2009-10, the company tax forecast on an income-year basis was closely aligned
with the GOS forecast (the red and blue bars in the left-hand panel of Figure 4.19). However, the
outcomes for the two series diverged significantly (the green and blue dots in the left-hand panel of
Figure 4.19). GOS grew in total by 13.9 per cent over these two years, whereas company tax
(income-year) fell in total by 8.9 per cent.
A major contributing factor to the weak outcome for company tax was high depreciation deductions
(see Figure 4.20 below), relating to the huge increase in capital investment as part of the mining
boom, and higher royalty expenses (also related to increased mining activity). As mentioned, GOS is
measured gross of depreciation expenses. At the time, depreciation expenses were forecast separately
in the company tax model based on historical trends. The methodology for forecasting depreciation
deductions has been improved in light of this experience, and now utilises the investment forecasts
produced by the macroeconomic forecasters.
Figure 4.20: Depreciation and Other Investment-related Deductions
60
Per cent of net taxable income
Per cent of net taxable income
60
50
50
40
40
30
1999-00
2000–01 2001–02
2002-03
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
30
2009-10
Note: Deductions include depreciation, rents and royalties, research and development, mining exploration, capital
expenditure, capital allowances and industry payments.
59
Review of Treasury Macroeconomic and Revenue Forecasting
In addition, the timing adjustments made to the income-year forecast for 2009-10 proved to be
inaccurate (as shown in the right-hand panel of Figure 4.19). Company tax was expected to fall by
less on a cash, than on an income-year, basis, due to backwards-looking elements in the company tax
payment system. In the event, company tax fell by around the same amount in 2009-10 on both a cash
and income year basis. In fact, the company tax forecast for the 2009-10 income-year for a fall of
around 10 per cent turned out to be reasonably accurate (left-hand panel of Figure 4.19), even though
the cash forecasts for 2009-10 and 2010-11 were inaccurate. This, again, indicates problems with the
assumptions made in relation to the timing of cash payments of company tax.
The majority of company tax is paid in quarterly instalments, based on an instalment rate derived
from the most recently assessed tax return. This was expected to lead to overpayments of company tax
on a cash basis in 2009-10, since it was a year of declining taxable profits. However, as companies
became aware that the quarterly instalment payments that they were making in 2009-10 were likely to
overestimate their yearly tax liability, they elected to vary down their instalments payments (to a
greater extent than was anticipated in the forecasts).
Company tax revenue growth was also overestimated in 2010-11 and, to a lesser extent, in 2011-12.
The reasons behind these forecast errors will be difficult to discern until tax return data are available
for these income years6. In both years, the company tax growth rates were adjusted upwards in
translating from income-year to cash forecasts. This is based on the typical pattern of company tax
payments after a downturn, where instalments rates are varied upwards as profits recover7.
The counterfactual exercise finds that the forecast errors do not improve when the GOS outcomes are
applied to the current company tax model (as shown in the right-hand panel of Figure 4.16, above). In
particular, the overestimation of company tax increases in 2009-10, due to the GOS outcome being
significantly higher than forecast, with flow-on implications for the 2010-11 forecast error, due to the
lags in the company tax payment system.
Figure 4.21, below, shows the historical mismatch between estimates of company tax (income-year)
based upon GOS and actual company tax (income-year) outcomes. This shows that even when all of
the economic and tax schedule data are known, it is not possible to completely reconcile the actual
company tax outcomes with GOS. That said, the relationship was stable over the five-year period
between 2003-04 and 2007-08. In recent years the relationship appears to have deteriorated.
Figure 4.21: Actual Company Tax Collections versus Company Tax Calculated from GOS
1.05
Ratio
Ratio
1.05
Based on preliminary
Taxation Statistics data
1.00
1.00
0.95
0.95
0.90
0.90
0.85
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
0.85
2010-11
Note: The ratio is set equal to one in 2003-04.
6 The Taxation Statistics are released with a significant lag, with the 2009-10 edition being released on 1 May 2012.
7 See the box on page 5-22 of the 2012-13 Budget Paper No 1 for further discussion of this issue.
60
Section 4: Case Studies
It is possible that future revisions to the GOS outcomes may improve the relationship between GOS
and company tax in recent years. In the past, GOS has been revised significantly: for example, the
outcome for growth in GOS in 2008-09 was revised upwards from 6.8 per cent, as first published, to
14.2 per cent (as at the June 2012 National Accounts). As they stand, the current GOS outcomes for
2008-09 and 2009-10 are particularly difficult to reconcile with the company tax outcomes through
the GFC.
Significant work has been undertaken to improve the company tax model in response to the forecast
errors that have been observed in recent years. In particular, Treasury is developing a three sector
company tax model which splits the economy into mining, finance and insurance and other sectors.
This should allow the model to take better account of the different characteristics of these sectors,
such as:
•
the greater role of the mining sector in the Australian economy and its capital-intensive nature;
•
the measurement of the income of the finance sector in the National Accounts; and
•
substituted accounting periods, which makes it more difficult to accurately estimate the timing
of the receipt of cash payments of corporate tax. In particular many large financial companies
operate on an accounting year ending in September and many large mining companies operate
on an accounting year ending in December.
Notwithstanding these developments, due to the difficulty that Treasury has had forecasting revenue,
in particular company tax revenue, the Review recommends that:
Recommendation 9:
Treasury should give further consideration to the appropriate balance between the top-down versus
bottom-up approaches to forecasting revenue.
Recommendation 10:
Treasury, in conjunction with the ABS as necessary, should explore further ways of improving the
current methodology for forecasting corporate tax, and also consider alternatives to the current
methodology, which could perhaps be used to complement existing approaches.
GST
During the GFC period, the outcomes for GST revenue diverged significantly from the outcomes for
its main economic base (consumption subject to GST), which presented challenges for forecasting.
GST revenue fell in 2008-09, following the onset of the GFC (the light blue line in Figure 4.22). At
the time, growth in consumption subject to GST fell only marginally (the blue line). This series has
since been revised (the red line) and now better tracks GST revenue growth during 2008-09.
However, a large discrepancy opened up again in 2011-12 between GST revenue and consumption
subject to GST growth, with the National Accounts data again indicating a stronger economy than
would be suggested by the GST revenue data.
61
Review of Treasury Macroeconomic and Revenue Forecasting
Figure 4.22: GST and Consumption Subject to GST
8
Per cent
Per cent
8
GST revenue
6
4
6
4
Consumption subject to GST
(Decem ber 2009 National Accounts)
2
2
Consumption subject to
GST (June 2012
National Accounts)
0
0
-2
-2
-4
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
-4
2011-12
When the consumption subject to GST outcomes are applied to the GST model, the forecast errors
reduce significantly (as shown in the right-hand panel of Figure 4.16 above, the green bars), except in
2011-12. The consumption subject to GST outcome for 2011-12 may be revised in the future, as the
ABS reconciles the National Accounts consumption data with the ATO tax collections data.
Capital gains tax
The extent of the run-up in CGT revenue during Mining Boom Mark I was unanticipated, as was the
extent of the drop-off following the GFC. As shown in Figure 4.23 below, CGT peaked as a share of
revenue in 2007-08 at 6.3 per cent of revenue in that year. The stock market began to decline in
December 2007 and by 2010-11 CGT comprised only 2.3 per cent of taxation revenue.
Figure 4.23: Evolution of Capital Gains Tax Forecasts (as a Percentage of Taxation Revenue)
7
Per cent
Per cent
6
6
2008-09
Budget
2010-11
Budget
5
4
3
2009-10
Budget
2
2011-12
Budget
2012-13
Budget
1
62
5
4
3
0
1992-93
7
2
1
1995-96
1998-99
2001-02
2004-05
2007-08
2010-11
0
2013-14
Section 4: Case Studies
The CGT forecast errors reduce significantly when actual asset price outcomes are applied to the
current model, (as shown in the right-hand panel of Figure 4.16, the light blue bars), indicating that if
asset prices had been known at the time the forecasts were being prepared, the current CGT model
would not have generally overestimated revenue over this period.
This observation leads the Review to recommend that:
Recommendation 11:
The technical specialist with deep financial market experience employed by Treasury should be
tasked with improving the accuracy of the technical assumptions for equity and housing prices that
are used to generate the capital gains tax revenue forecasts.
63
Review of Treasury Macroeconomic and Revenue Forecasting
APPENDIX A: REVENUE FORECASTING MODELS
Current Company Tax Model (Figure 4.24)
Of the National Accounts variables forecast by DED, company tax is most closely related to gross operating surplus (GOS), which is a measure of company
profits. However, the relationship is far from simple. There are many conceptual differences between GOS and corporate taxable income, including
depreciation, the treatment of losses, net interest income and capital allowances. There are also significant timing differences between profits being earned
and the receipt of the associated tax revenue. As a rule-of-thumb, around 60 per cent of company tax is received in the year the profit is generated, with the
remaining 40 per cent received in the following year. This means that at times when profits are varying significantly from year-to-year, as occurred over the
GFC period, incorrect assumptions around timing can be a major source of forecast error.
Part A, of Figure 4.24, below, shows the steps undertaken to translate from GOS (economic income on a National Accounts basis) to gross operating
profit/loss subject to tax (economic income on a tax basis). Part B shows how deductions and other items are subtracted, and the average tax rate is applied, to
generate company tax on an income-year basis (the forecasts for which are shown in the left-hand panel of Figure 4.19, the blue bars). Part C shows how
timing adjustments are applied, and government policy measures are taken into account, to generate company tax on a cash basis (the forecasts for which are
shown in the right-hand panel of Figure 4.19, the red bars).
Capital Gains Tax Forecasting Model — Superannuation Funds (Figure 4.25)
Broadly, the first step in forecasting CGT payable by superannuation funds is to estimate the stock of unrealised gains and losses, as at the end of the previous
year. Unpublished APRA and ATO data are used to assist in this process. Since these data are generally released with a lag, modelled estimates of capital
gains and losses are used to build the stock from the last available data point.
The second step is to forecast the current year capital gains. Unpublished APRA data is used to estimate the level and mix of assets held by superannuation
funds, and data on year-to-date price movements in the ASX 200 and house prices are used to calculate asset price growth in the current year. Beyond the
current year, all asset prices are assumed to grow in line with nominal GDP. Current year losses are assumed to be a certain fraction of current year gains,
based on recent historical averages.
A realisation rate is then applied to the capital gains (and an application rate to the losses), to account for the fact that the assets may not be sold in the current
year and that losses from previous years may be applied to the current year. The realisation rate is based on recent historical averages, unpublished data from
the ATO and judgement.
The final steps are to apply the CGT discount, and then multiply the net capital gain by the effective tax rate (15 per cent in the case of superannuation funds)
to calculate a CGT forecast.
64
Section 4: Case Studies
Figure 4.24: Current Company Tax Model
65
Review of Treasury Macroeconomic and Revenue Forecasting
Figure 4.25: Capital Gains Tax Forecasting Model — Superannuation Funds
66